Pyruvate Oxidase Is a Determinant of Avery's Rough Morphology

Abstract

In pioneering studies, Avery et al. identified DNA as the hereditary material (A. T. Avery, C. M. MacLeod, and M. McCarty, J. Exp. Med. 79:137-158, 1944). They demonstrated, by means of variation in colony morphology, that this substance could transform their rough type 2 Streptococcus pneumoniae strain R36A into a smooth type 3 strain. It has become accepted as fact, from modern textbook accounts of these experiments, that smooth pneumococci make capsule, while rough strains do not. We found that rough-to-smooth morphology conversion did not occur in rough strains R36A and R6 when the ability to synthesize native type 2 capsule was restored. The continued rough morphology of these encapsulated strains was attributed to a second, since-forgotten, morphology-affecting mutation that was sustained by R36A during strain development. We used a new genome-PCR-based approach to identify spxB, the gene encoding pyruvate oxidase, as the mutated locus in R36A and R6 that, with unencapsulation, gives rise to rough colony morphology, as we know it. The variant spxB allele of R36A and R6 is associated with increased cellular pyruvate oxidase activity relative to the ancestral strain D39. Increased pyruvate oxidase activity alters colony shape by mediating cell death. R36A requires a wild-type spxB allele for the expression of smooth type 2 morphology but not for the expression of smooth type 3 morphology, the phenotype monitored by Avery et al. Thus, the mutated spxB allele did not impact their use of smooth morphology to identify the transforming principle.

The bacterium Streptococcus pneumoniae owns the important distinction of having introduced the world to the concept that DNA is the hereditary material. In their landmark studies, Avery, MacLeod, and McCarty identified DNA as the substance that could convert or transform one type of S. pneumoniae into another (1). The indicator of the transformation event was the ability of an unencapsulated strain of S. pneumoniae with rough colony morphology (R) to acquire a new encapsulated smooth (S) appearance (1).

The R and S nomenclature has long been used to describe the colony morphology of S. pneumoniae: R colonies appear rough and irregular on solid media, while S colonies appear smooth and shiny. In the celebrated work of Griffith, these terms were originally used to correlate colony morphology with virulence properties (4). Later, R and S came to describe the encapsulation state of the bacterium because it was found that, in contrast to S colonies, R colonies do not produce type-specific polysaccharide capsule. The R designation is actually an umbrella term that has been used to describe many morphologically distinct unencapsulated S. pneumoniae mutants. Although all strains producing R colonies have lost their ability to make capsule, not all R colonies look the same, because the strains in question may have accumulated additional mutations that further alter colony morphology.

One R variant that has sustained multiple colony morphology-altering mutations is S. pneumoniae R36A, the strain used by Avery et al. to demonstrate transformation (1). The history of R36A indicates that the R phenotype is the result of at least two independent mutational events (1, 9). R36A is a descendant of S. pneumoniae R36, a derivative of the smooth type 2 capsule-producing strain S. pneumoniae D39 (Fig. 1). R36 was isolated by serially passaging D39 in broth containing pneumococcal type 2 antiserum a total of 36 times (1, 9). The resulting strain was both unencapsulated and easily transformed (1, 9). However, the phenotype of R36 was unstable: bacterial cultures grown in blood broth would become populated with morphological variants that had lost the ability to be transformed (1, 9).

FIG. 1.

Lineage of S. pneumoniae laboratory strains derived from D39 and their associated phenotypes and genotypes.

In order to isolate a strain with a fixed phenotype, the morphological variants of R36 were individually examined for their ability to be transformed. The strain eventually chosen by Avery's group for further work was R36A, a morphological variant that retained the transformation properties of the parental R36 strain but did not dissociate further into undesirable forms (1, 9) (Fig. 1). The newly isolated R36A strain has been used ever since for the study of pneumococci and is the progenitor of the S. pneumoniae Rx and R6 lineages, which account for the bulk of S. pneumoniae laboratory strains used today (5, 15) (Fig. 1). Later genetic analysis revealed that the loss of encapsulation in R36A and, presumably, R36, is due to a deletion in the gene cluster responsible for the biosynthesis of type 2 capsule (6). The morphology-changing event beyond the loss of encapsulation that occurred in R36A during strain development has not been defined.

The S. pneumoniae spxB gene encodes pyruvate oxidase, an enzyme that decarboxylates pyruvate in a four-step reaction that yields acetylphosphate, hydrogen peroxide, and carbon dioxide at the expenditure of oxygen (16). It has been shown that SpxB-generated acetylphosphate is an important source of ATP for S. pneumoniae during aerobic growth and during times of oxidative stress (13). The spxB gene is not strictly essential for growth in vitro (16), but its importance in the biology of S. pneumoniae has been demonstrated in a number of studies. Strains lacking spxB exhibit a loss of adherence and virulence (16) and are more susceptible to killing by hydrogen peroxide (13). Intrastrain colonial variation is associated with changes in pyruvate oxidase expression, though such variation is not directly attributable to SpxB (10, 13, 19). The so-called opaque and transparent phase variation phenomenon differs from the R and S phenomenon in that opaque and transparent variants are isolated under different growth conditions and phase variation occurs independently of the encapsulation state of the cell (10). The hydrogen peroxide by-product of pyruvate oxidase may help S. pneumoniae create niches for itself in the respiratory tract by killing competing bacterial pathogens (12) and may drive genetic change in this organism (11).

Described here is the identification of spxB as the mutated gene in R36A and its descendant R6 that, in addition to loss of encapsulation, gives rise to rough colony morphology as we know it. The physiological ramifications of the mutation, possible mechanisms for rough colony morphology development, and relevance to Avery's original experiments are also discussed.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

See Table 1 for the S. pneumoniae strains used in these studies. The S. pneumoniae strains were routinely propagated in Todd-Hewitt broth containing 0.5% yeast extract (Difco, Detroit, Mich.) or on TSA II agar containing 5% sheep blood (BBL, Cockeysville, Md.) at 37°C with 5% CO₂. When required, erythromycin was added to agar plates at a concentration of 0.3 μg/ml with nutrient broth (Difco) overlays containing 0.8% Bacto agar (Difco). When appropriate, blood agar plates were supplemented with 3,900 or 7,800 U of catalase.

TABLE 1.

S. pneumoniae strains used in this study

Strain	Relevant characteristics	Source or reference(s)
D39	Type 2 strain, cps2+spxB+	1
R36A	D39 Δ(cps2 2538-9862) spxB	1, 6
R6	R36A with increased transformation efficiency	5
6303	Type 3 strain	American Type Culture Collection
AB1	D39 ermB inserted between cps2CD, in frame	This study
AB2	R6 transformed with D39 cps2ABCermDEFGH amplicon	This study
AB7	AB2 transformed with AB28 genomic DNA	This study
AB8	R6 ΔspxB::ermB	This study
AB9	D39 ΔspxB::ermB	This study
AB10	R36A transformed with D39 cps2ABCermDEFGH amplicon	This study
AB11	AB10 transformed with D39 spxB+ amplicon	This study
AB12	R36A ΔspxB::ermB	This study
AB13	R36A transformed with D39 spxB+ amplicon	This study
AB14	R6 transformed with D39 spxB+ amplicon	This study
AB15	Type 3 encapsulated strain of R36A	This study
AB28	D39 Δcps2::ermB	This study

Genetic manipulations and strain construction.

S. pneumoniae genomic DNA was isolated according to previously published methods (2). The PCR-based ordered genomic library was constructed as described previously (2) except that the DNA was amplified under high-fidelity conditions with Platinum Taq High Fidelity polymerase (Invitrogen, Carlsbad, Calif.). Routine transformations were performed as described (2) except that the concentration of DNA in the transformation mix was 1 μg/ml. Gene insertions, deletions, and substitutions were accomplished by transforming S. pneumoniae with DNA amplicons generated with Platinum Taq High Fidelity polymerase under the PCR conditions suggested by the manufacturer. See Table 2 for the amplicons and PCR primer sequences used in strain construction.

TABLE 2.

Amplicons and oligonucleotide primers used in this study

Amplicon	Primer	Sequence (5′ → 3′)
cps2-erm	cpsCupF	AATTCTCAAGATTTAAAAGGGACAGGTC
	cpsCupR	CCTTCTTAATTCTATTTCATTTTGTCCAAATCTGGAAC
	cpsermF	AAATGAAATAGAATTAAGAAGGAGTGATTACATGAACAAAA
	cpsermR	CTTCCTCCTCGACTCATAGAATTATTTCCTCCCGTTAAA
	cpsDdnF	TATGAGTCGAGGAGGAAGTTATGCCAACGTTAGAAA
	cpsDdnR	CAATAATCCGCTTAGCAATTACATGAC
cps2ABCermDEFGH	cps2F	TCTTAGTTCCATGGGATGCTTTCTG
	cps2R	AACTGTAATACGACCGAAAGCTTCTTTT
cps2 knockout	cpsIIKOa1	GGATCCCGGGACAGTAGGGG
	cpsIIKOa2	AACGGGAGGAAATAACCTGAGTACAATCC
	cpsIIKOb1	GGATTGTACTCAGGTTATTTCCTCCCGTT
	cpsIIKOb2	GTAAGCGCCCAATAAAGAAGTTATGGAAATAAG
	cpsIIKOc1	CTTATTTCCATAACTTCTTTATTGGGCGCTTAC
	cpsIIKOc2	GCAAATTCAGCCCGGGCTTTTTC
spxB+	spxBF	ATCTTTACTGTGCCTCTTGTGATTCTCATG
	spxBR	TGGTCTACATCCTCAACCTCGATATGAATA
spxB knockout	spxBKOupF	TCATTGGCTCAACCATCAATCATAGTGGAA
	spxBKOupR	CCTTCTTAATTCAGATGCAGTAATTTTCCCTTGAGTCATAA
	spxBKOermF	CTGCATCTGAATTAAGAAGGAGTGATTACATGAACAAAA
	spxBKOermR	AAGGCTTTGACGACTCATAGAATTATTTCCTCCCGTTAAA
	spxBKOdnF	TATGAGTCGTCAAAGCCTTCAAGGAAAAATACGAAGCAG
	spxBKOdnR	TTCAAGAAACACGTGGTGGTCGTAAC

The ability to synthesize type 2 capsule was restored to R36A and R6 as follows. The AB1 strain was constructed by inserting an erythromycin resistance gene between the cps2C and cps2D genes of D39 in such a way that the operon structure cps2ABCD was not disrupted. The cps2-erm amplicon was generated by fusion PCR (8) so that D39 cps2C and upstream DNA and cps2D and downstream DNA each flanked the pAMβ1 ermB gene. Primers were designed from published DNA sequences for the D39 cps2 gene cluster (GenBank accession number AF026471) and the ermB gene of pAMβ1 (GenBank accession number Y00116). D39 was transformed with the cps-erm amplicon, and transformants were selected on medium containing erythromycin. Transformants containing the desired insertion of ermB were identified by diagnostic PCR analyses. The retention of parental morphology and type 2 capsule production by AB1 was confirmed by visual inspection and a positive Quellung reaction. Genomic DNA from AB1 was used as the template in a PCR to create the cps2ABCermDEFGH amplicon. This amplicon was then used to transform R6 and R36A to erythromycin resistance, creating AB2 and AB10, respectively. Strain confirmation entailed PCR analyses and immunological analysis via the Quellung reaction. The AB28 strain was constructed by transforming D39 with a fusion PCR amplicon containing DNA regions upstream of the cps2 gene cluster, ermB, and DNA regions downstream from cps2. Erythromycin-resistant transformants containing the cps2 deletion were verified by visual inspection and PCR analyses. The R6 S2 variant AB7 was generated by the transformation of AB2 with genomic DNA from AB28. Transformants were individually examined for the acquisition of colony morphology similar to that of D39. R36A and R6 strains containing the spxB⁺ allele, AB13 and AB14, respectively, were constructed by transforming the parental strains with a PCR amplicon containing the D39 spxB gene and then selecting for morphological variants. Primers were designed based on the published DNA sequence of R6 (GenBank accession number AE007317), and D39 genomic DNA was used as the PCR template. The spxB gene was deleted from the genome of R6, D39, and R36A with a fusion PCR amplicon containing the R6 DNA region upstream of spxB, ermB, and R6 DNA regions downstream of spxB to create strains AB8, AB9, and AB12, respectively. Erythromycin-resistant transformants were screened for the loss of spxB by PCR analysis. R36A S3 transformants, including AB15, were obtained with genomic DNA isolated from the S. pneumoniae type 3 strain 6303. Individual transformants were visually screened for those exhibiting the characteristic mucoid morphology.

Capsule production.

The production of type 2 capsule by the various pneumococcal strains was confirmed by the Quellung reaction performed with high-titer S. pneumoniae type 2 antiserum (Statens Serum Institut, Copenhagen, Denmark) at the dilution of 1:16. The amount of type 2 capsule produced by the various strains was determined via a colorimetric assay for the detection of uronic acids (3, 7). Bacterial pellets suitable for testing were prepared from cells grown to an optical density at 600 nm of 0.3 and washed with 1 volume of water. Standard curves were prepared from purified type 2 pneumococcal polysaccharide powder (American Type Culture Collection, Manassas, Va.).

Morphological variation.

S. pneumoniae transformants were plated at a density of 300 CFU on blood agar plates. After overnight incubation, approximately 3 × 10⁴ to 1 × 10⁵ individual colonies were visually examined for gross changes in morphology.

Pyruvate oxidase assays.

Pyruvate oxidase activity was determined with an assay that measures pyruvate-dependent liberation of hydrogen peroxide in S. pneumoniae cell lysates. Reactions contained 50 mM potassium phosphate, pH 6.5, 0.05 mM thiamine pyrophosphate, 0.0097 mM flavin adenine dinucleotide, 0.97 mM MgSO₄, 1.5 mM sodium pyruvate, 0.2 U of horseradish peroxidase per ml, and 100 μM Amplex Red reagent (Molecular Probes, Eugene, Oreg.). Reactions were run in a total volume of 200 μl with 10 μl of cell lysate. Bacterial cell lysates were prepared from cultures grown to an optical density at 600 nm of 0.3. Cell pellets were washed with 1 volume of 0.1 M sodium phosphate, pH 7.4, resuspended in 0.1 volume of buffer containing 0.1% Triton X-100, and incubated for 10 min at 37°C. Standard curves were generated with fresh dilutions of 30% hydrogen peroxide. Pyruvate oxidase assays were incubated at 37°C, and the absorbance of the reaction was read at 563 nm every 5 min for 1 h.

The release of hydrogen peroxide by S. pneumoniae cells was determined by a modification of the protocol provided in the Amplex Red hydrogen peroxide/peroxidase assay kit instruction manual (Molecular Probes). Briefly, cell pellets obtained from exponentially growing cultures were resuspended in 1 ml of Krebs-Ringer phosphate buffer to an optical density at 600 nm of 0.1. A 20-μl aliquot (≈10⁶ CFU) of each cell suspension was added to 100 μl of a solution containing 50 μM Amplex Red reagent and 0.1 U of horseradish peroxidase per ml in Krebs-Ringer phosphate that was prewarmed to 37°C. The reaction was incubated at 37°C, and the absorbance was read at 563 nm every 15 min for 1 h.

Colony viability.

Individual bacterial colonies were harvested with sterile toothpicks, and the cells were dispersed in broth. The cell suspension was serially diluted, and aliquots were spread on blood agar. The plates were incubated overnight, and the resulting colonies were enumerated.

Nucleotide sequence accession numbers.

The spxB nucleotide sequences of the strains used in this study have been deposited in GenBank and assigned the accession numbers AY254851 (R36A), AY254852 (D39), AY254853 (AB7), and AY254854 (AB15).

RESULTS

Type 2 capsule alone does not determine S2 morphology.

Classic S. pneumoniae nomenclature designates strains exhibiting rough colony morphology as R forms and strains exhibiting smooth colony morphology as S forms. The S designation is further amended with a number to denote the type of capsule that is made by the smooth strain. For example, smooth strains expressing type 2 capsule are referred to as S2, and smooth strains expressing type 3 capsule are referred to as S3. We have adhered to this nomenclature system here. The R and S phenotypes specifically describe the colonial morphology of S. pneumoniae grown on blood-containing agar (1); the same culture conditions were used in these studies.

Popular thought holds that the restoration of capsule synthesis to R strains should result in S colony morphology. The loss of encapsulation in R36A and its descendants is the result of a deletion in the cps2 gene cluster that eliminates the expression of half of the gene products necessary for type 2 capsule biosynthesis (6). We performed a targeted replacement of the deleted cps2 region in the R36A derivative R6 to create strain AB2 (Fig. 2). Although AB2 was shown immunologically to elaborate type 2 capsule on the cell surface by the Quellung reaction (data not shown), it did not regain the S2 morphology of its ancestor D39 (Fig. 3A, 3B and 3C). An R36A strain with restored type 2 capsule biosynthesis, AB10, behaved similarly (data not shown). A more precise quantitative method revealed that the rough AB2 strain produced near wild-type levels of type 2 capsule, approximately 87% that made by D39 (Table 3).

FIG. 2.

Generation of encapsulated S. pneumoniae strains AB2 and AB10. R36A and R6 were transformed with a PCR product encompassing the portion of the D39 type 2 capsule biosynthetic gene cluster that was lost during the isolation of the ancestral strain R36 from D39. Transformants that had reacquired the missing cps2 region by allelic exchange were identified by selection with erythromycin and immunological testing for type 2 capsule biosynthesis. The crosshatched box in the R36A and R6 cps2 gene cluster depicts a fusion of portions of the cps2A and cps2H genes that occurred during the deletion of the cps2 DNA. The white arrow in the D39 cps2ABCermDEFGH PCR product denotes the ermB marker that was placed between cps2C and cps2D to monitor allelic exchange. The polka-dotted arrows in the cps2 gene cluster of AB2 and AB10 represent hybrid cps2A and cps2H genes that were created during allelic exchange.

FIG. 3.

Colony morphology of various S. pneumoniae strains grown on blood agar. (A) D39; (B) R6; (C) AB2; (D) AB7; (E) AB28; (F) R36A; (G) AB14; (H) AB15.

TABLE 3.

Type 2 capsule production of various strains

Strain	Morphology	% Type 2 capsule relative to D39a
D39	S	100
R6	R	0
AB1	S	73
AB2	R	87
AB7	S	88
AB28	R	0

^aThe results reported are averages of three assays.

We found that a conversion from R to S2 morphology could occur in AB2 at a frequency of 10⁻³ per CFU when this strain was transformed with genomic DNA from D39 to produce the R6 S2 variant AB7 (Fig. 3D). The morphology change was reversible because transformation of AB7 with genomic DNA from R6 resulted in the isolation of R colonies at approximately the same genetic frequency. The frequencies at which R to S and S to R morphology changes took place were consistent with the allelic exchange of a single marker (2). The unknown gene associated with the new S2 phenotype was not linked to the cps2 gene cluster because DNA derived from the AB28 strain still mediated the morphology change. Also, the S2 morphology was not due to increased type 2 capsule production because AB7 made essentially the same amount of capsule as its parental strain AB2 (Table 3).

spxB is a determinant in type 2 colony morphology.

To gain insight into why the type 2 encapsulated strains of R36A and R6 had retained rough morphology, we carefully read the original transformation paper by Avery et al. (1). The account of the derivation of R36A revealed that the rough colony morphology of this strain was actually the result of two mutational events: the first led to loss of capsule, and the second further altered colony appearance. We speculated that the second unknown mutation in R36A, and by extension R6, not only contributed to the unique rough morphology of these strains but also prevented the development of smooth morphology even in the presence of capsule.

To identify the unknown gene that mediates colony morphology, we screened a PCR-based ordered genomic library of R6 (2) for a gene that could confer R morphology back to AB7 when acquired by allelic exchange. The PCR-based ordered library was comprised of 522 overlapping PCR amplicons of approximately 4 kb that represented the entire genome of S. pneumoniae R6. This library was similar to one that we had previously used to identify drug targets in S. pneumoniae R6 except that the amplicons were generated under high-fidelity PCR conditions rather than error-prone PCR conditions (2).

Pools of 60 and then six individual contiguous PCR amplicons were tested for their ability to confer R morphology to AB7 (Fig. 4, step 1 and step 2). Once a positive pool of six amplicons had been identified, the individual amplicons contained within that pool were tested for the ability to confer R morphology (Fig. 4, step 3). This approach led to the identification of amplicon 165 as the one containing a gene that mediated morphological switching. Examination of the genetic content of amplicon 165 (Fig. 4, step 3) and subsequent overlapping amplicon analysis (Fig. 4, step 4) revealed that the R6 spxB gene confers R morphology to AB7. The spxB gene encodes pyruvate oxidase, an enzyme that converts pyruvate to acetylphosphate, hydrogen peroxide, and carbon dioxide at the expense of oxygen (16). Subsequent transformation of the AB2 and AB10 strains with spxB-containing PCR products amplified from genomic DNA isolated from AB7 and D39 produced colonies exhibiting S2 morphology (data not shown).

FIG. 4.

Identification of the S. pneumoniae R6 gene that confers R morphology back to the R6 S2 variant AB7. An ordered library of PCR products that represent the entire R6 genome was screened for a single amplicon that could restore R morphology back to AB7 by allelic exchange. To facilitate screening, pools comprising either 60 or six sequential amplicons representing contiguous regions of the genome were tested first. Amplicon 165 was found to confer R morphology to AB7 and contains two RUP (repeat unit of pneumococcus) genetic elements and four complete open reading frames: spxB, encoding pyruvate oxidase; hypo, encoding a hypothetical protein; and two tnp, encoding transposases. spxB was identified as the gene that mediates colony morphology by testing a new overlapping amplicon that contained only this gene as a complete open reading frame.

DNA sequence analysis of R36A, R6, D39, and AB7 revealed the existence of two spxB alleles. One, presumably the wild-type allele, was found in D39 and AB7 (spxB⁺) and was associated with S2 morphology. The other allele was present in R36A and R6 (spxB) and was associated with R colony morphology. The different phenotypic outcomes of the two alleles were attributed to the mutation in spxB; relative to spxB⁺, spxB contains two amino acid substitutions, N95D and T282A, in its coding sequencing. DNA sequence analysis was also performed on the spxB promoter region of R36A, R6, D39, and AB7, and no sequence differences were found among the four strains.

In addition to encapsulated R strains, the spxB⁺ allele also mediated a colony morphology change in the parental R36A and R6 rough strains. AB13 and AB14 (Fig. 3G) colonies appeared smooth but were not as large as their S2 counterparts (Fig. 3D). The colony morphology of these strains was similar to that exhibited by AB28, an unencapsulated strain of D39 (Fig. 3E).

spxB is associated with increased pyruvate oxidase activity.

In preparation for cellular pyruvate oxidase studies, we constructed spxB null strains of R36A, R6, and D39. We noted that AB8 and AB12 exhibited altered colony morphology on blood medium, one that was similar to that of AB13, AB14, and AB28 (data not shown). These results suggested that the inability to form smooth colonies and the characteristic rough morphology of R36A and R6 might be due to an increase in cellular SpxB activity relative to that of D39.

Pyruvate oxidase activity was measured in cell lysates of the various strains with an assay that monitors the pyruvate-dependent generation of hydrogen peroxide. Activity was readily detected in cell lysates prepared from R36A and R6, 38.6 ± 2.8 and 53.3 ± 5.6 pmol of H₂O₂ min⁻¹, respectively, but not in lysates made from D39, any of the spxB null strains, or AB7. Therefore, to demonstrate more definitively the differences in pyruvate oxidase activity among the S. pneumoniae strains, we measured the SpxB-associated release of hydrogen peroxide from the various strains. The results of these experiments concurred with those of the enzyme assays: R36A and R6 exhibited a marked increase in pyruvate oxidase activity relative to spxB⁺ and spxB null strains. R36A and R6 cells released approximately three to four times more hydrogen peroxide than did D39 and AB7 (Table 4). Taken together, our genetic and biochemical results indicated that the spxB allele in R36A and R6 was associated with increased cellular pyruvate oxidase activity.

TABLE 4.

Hydrogen peroxide release assays

Strain	Genotype	Avg H2O2 released (pmol/min)a ± SE
D39	spxB+	25 ± 4
R36A	spxB	63 ± 7
R6	spxB	95 ± 3
AB7	spxB+	15 ± 3
AB8	ΔspxB::ermB	20 ± 3
AB9	ΔspxB::ermB	13 ± 5
AB12	ΔspxB::ermB	17 ± 11
AB15	spxB	129 ± 2

^aResults are averages of three assays.

Colony viability is a determinant of morphology.

Unencapsulated strains of S. pneumoniae are considered rough because, upon prolonged incubation, the colonies collapse and become ruffled in appearance. The AB28 colonies (Fig. 3E), which are the result of an overnight incubation, would eventually assume this appearance. After a comparable incubation period, R36A and R6 colonies appeared nothing like those of AB28. The R36A and R6 colonies were similar in diameter to those of the unencapsulated D39 strain but exhibited a flattened umbilicated appearance after overnight incubation (Fig. 3B and Fig. 3F). One way to interpret this unique appearance of R36A and R6 is that initially, the colony grows normally and spreads outward, but then cell death occurs in the center. To test this hypothesis, we measured the number of viable CFU per colony for a variety of strains. The R36A and R6 R colonies were found to contain 70 times fewer viable CFU than the S colonies of D39 (Table 5). The decreased number of viable CFU found in R36A and R6 colonies could, in part, be attributed to the spxB allele, because the deletion of the gene caused colony CFU counts to rise 10-fold relative to that of the parental strains (Table 5). Also, the replacement of the spxB allele with the spxB⁺ allele in both R36A and R6 resulted in an increase in the number of viable colony CFU/original colony by some 30-fold (Table 5).

TABLE 5.

Colony viability of various strains

Strain	Genotype	Avg. CFU per colony (104)a ± SE
D39	spxB+	70 ± 17
R36A	spxB	1 ± 0.2
R6	spxB	1 ± 0
AB8	ΔspxB::ermB	16 ± 4
AB12	ΔspxB::ermB	12 ± 3
AB13	spxB+	28 ± 10
AB14	spxB+	28 ± 1
AB15	spxB	37 ± 12

^aThe results are averages of five colonies.

Since spxB was associated with both increased pyruvate oxidase activity and hydrogen peroxide release, we considered the possibility that high concentrations of hydrogen peroxide cause the cell killing. Hydrogen peroxide specifically kills S. pneumoniae by inactivating essential cellular enzymes and depleting ATP pools, which are partially generated by acetate kinase from SpxB-derived acetylphosphate (13). Unlike other bacteria, such as Escherichia coli, S. pneumoniae does not make catalase, so high levels of endogenously made hydrogen peroxide can accumulate. The sheep red blood cells incorporated into the TSA growth medium used for S. pneumoniae are meant to counteract the negative effects of this hydrogen peroxide. Alternatively, exogenous catalase may be added to the growth medium in place of blood to facilitate the growth of the bacterium. If the decreased viability counts and the R morphology of R36A and R6 were caused by excess hydrogen peroxide, then growth on blood medium additionally supplemented with catalase should have caused the colony morphology of R36A and R6 to change. However, we did not observe any colonial morphology change when R36A and R6 were grown on blood plates in the presence of catalase.

S3 morphology does not require spxB⁺.

In the paper in which the identification of the transforming principle was described, Avery's laboratory had monitored transformation by examining the ability of R36A to express the S3 morphology of a type 3 capsule-producing S. pneumoniae strain and not the S2 morphology of its ancestor D39 (1) (Fig. 3F and Fig. 3H). This choice was apparently made early on because S3 transformants of the parental R36 strain were more readily obtained than the transformants of other capsule types (9). Our work suggested that in order to regain S2 morphology, R36A and R6 must reacquire both the missing capsule biosynthetic genes and spxB⁺ during transformation. We wondered how the spxB allele would impact the ability of R36A to express the S3 phenotype. In isolating S3 transformants, was Avery's laboratory selecting for a double genetic event?

To address this question, we isolated S3 transformants of R36A (Fig. 3H) and then determined what spxB allele they contained. Genetic analysis of several R36A S3 transformants revealed that S3 morphology did not require the acquisition of a new allele of spxB because spxB was retained in these strains. Consistent with this finding, we determined that S3 transformants were obtained at similar frequencies (≈10⁻⁴/CFU) regardless of whether R36A contained spxB or spxB⁺. The R36A S3 transformant AB15 was found to release copious amounts of hydrogen peroxide, as would be expected for a strain bearing an spxB allele (Table 4). However, the number of viable CFU in the R36A S3 colony was similar to that seen with strains harboring the spxB⁺ allele (Table 5).

DISCUSSION

In this work, we identified the spxB gene as a variant locus that acts in concert with the loss of type 2 encapsulation to produce the unique R colony morphology of S. pneumoniae R36A and R6. Our genetic and morphological data make a compelling argument that the mutation of spxB is the morphology-changing genetic event selected by Avery et al. (1) during the isolation of R36A from R36. R36A and its direct descendant, R6, exhibited characteristic rough morphology when they contained the spxB allele but assumed a morphology that was consistent with an unencapsulated D39 strain when spxB was replaced by the spxB⁺ allele. The spxB⁺ allele was also required by encapsulated strains of R36A and R6 to achieve the full S2 morphological phenotype of the D39 ancestral strain. It would seem that the growth conditions used by Avery's group favored the genesis and proliferation of a strain like R36A with higher SpxB activity and increased hydrogen peroxide production. R36A was isolated from R36 that had been cultured aerobically in broth containing blood, an agent that facilitates the decomposition of hydrogen peroxide. Mutants with increased pyruvate oxidase activity, increased hydrogen peroxide production, and, presumably, an enhanced ability to generate ATP under aerobic conditions could easily have arisen and been selected for under such conditions. It is also noteworthy that after R36A was isolated, a method of treating beef heart infusion broth with activated charcoal, a hydrogen peroxide-decomposing material, was adopted by Avery's laboratory to ensure consistent growth and transformation of R36A (9). This line of reasoning could have been solidified if the phenotype and genotype of R36 could have been compared with those of D39, R36A, and R6. Regrettably, R36 was not available from either public or private sources and, indeed, may represent a strain that has been lost over time. The finding that spxB does not interfere with the expression of an S3 phenotype and, thus, would have in no way hampered the efforts to identify the transforming principle lends further credence to our hypothesis that this alternative spxB allele arose in Avery's laboratory.

The data presented here suggest that the rough morphology of R36A and R6 is in part the result of a loss of viability of bacterial cells within the colony that is caused by increased pyruvate oxidase activity. Why an increase in pyruvate oxidase activity would result in cell death remains to be determined. The simplest explanation is that the excess hydrogen peroxide produced as the result of the increased SpxB activity is toxic to the cells. The umbilicated appearance of R36A and R6 colonies is consistent with a hydrogen peroxide toxicity hypothesis: at the center of the colony, where the concentration of hydrogen peroxide would be the highest, cell death occurs, and at the periphery, where hydrogen peroxide concentrations are lower, cells remain viable. Our finding that further supplementation of the blood medium with catalase has no impact on colony morphology does not necessarily rule out this idea. It could be that the local concentration of hydrogen peroxide in the center of the colony is so great, or even unreachable, that no amount of supplementation would be sufficient to ameliorate the negative effects. The results that we obtained with the R36A S3 transformants are also consistent with the idea that a loss of viability of cells in the colony is due to hydrogen peroxide killing. S3 transformants make copious amounts of capsule, as evidenced by the large size and mucoid nature of the colonies. This excess capsular material could promote cell viability in the face of high levels of hydrogen peroxide by creating a buffer zone between the vulnerable intracellular targets of the bacterium and the hydrogen peroxide released by neighboring cells. In this case, the bacterium would only need to contend with its own endogenous hydrogen peroxide.

While a hydrogen peroxide toxicity hypothesis accommodates all of our data, it is not the only reason that cellular viability could be lost in the R36A and R6 colonies. Another possibility is that since pyruvate oxidase is an enzyme involved in central metabolism, increased activity of the enzyme could lead to preferential synthesis of acetylphosphate at the expense of NAD generation, which could contribute to cell death. Although attractive, this hypothesis does not easily explain our results with the R36A S3 transformants. It appears that further experimental work is needed to determine the precise mechanism by which increased pyruvate oxidase activity affects cellular viability in R36A and R6 colonies.

Our suggestion that the increased pyruvate oxidase activity of R36A and R6 leads to hydrogen peroxide-mediated killing appears to be inconsistent with a previous report in which pyruvate oxidase was identified as a hydrogen peroxide resistance determinant (13). If pyruvate oxidase confers hydrogen peroxide resistance, then it seems reasonable to assume that S. pneumoniae cells with higher pyruvate oxidase activity should suffer no loss of viability from the concomitant increase in hydrogen peroxide production; as levels of the toxic by-product increase, so too would the ability of the cell to resist its deleterious effects. Such reasoning would seem to invalidate our hydrogen peroxide toxicity model.

We believe, however, that the published data on pyruvate oxidase-mediated hydrogen peroxide resistance do not provide sufficient evidence to eliminate this model from a list of possible explanations for the spxB-associated loss of viability. In the previous report, it was concluded that pyruvate oxidase conferred resistance to hydrogen peroxide because cultures of D39 and R6x cells that expressed wild-type levels of the enzyme exhibited a 1% survival rate in the presence of high concentrations of exogenously added hydrogen peroxide, while cultures of D39 and R6x cells that did not express the enzyme exhibited a nearly 0% survival rate under the same conditions (13). The finding that R6 and type 4 phase variants that expressed 1.4- to 2-fold more pyruvate oxidase activity showed a 1% survival rate in the presence of high levels of exogenously added hydrogen peroxide and not the 0.05 to 0.1% survival rate of R6 and type 4 phase variants that expressed 1-fold pyruvate oxidase activity under the same conditions also contributed to this conclusion (13).

These results, in combination with others provided in the paper, clearly demonstrate that the ability to utilize SpxB-associated central metabolic processes could mean the difference between life and death for a small subpopulation of the cells in hydrogen peroxide-exposed cultures. However, the results presented in the paper did not definitively show that the cellular viability of S. pneumoniae remains constant under normal growth conditions even when pyruvate oxidase activity and hydrogen peroxide production increase manyfold, which would be required to effectively argue against the hydrogen peroxide toxicity model. Intuitively, the hydrogen peroxide protection power of pyruvate oxidase cannot completely counteract the toxicity of the hydrogen peroxide released as a result of its own enzymatic activity. If it did, then there would be no need to add catalase or blood to agar plates to facilitate the growth of S. pneumoniae in vitro. With no data to the contrary, the loss of viability observed in R36A and R6 could, as suggested by our model, be due to hydrogen peroxide killing.

The spxB gene has previously been implicated in the colony morphology differences that have been observed among some strains of S. pneumoniae grown in vitro. Encapsulated and unencapsulated pneumococci, including type 2 strains, can give rise to multiple morphological variants that differ in colony opacity (19). Some morphological variants that have a transparent appearance have been shown to express more SpxB than those that look opaque (13, 10), and at least one variant described as opaque produces a truncated inactive form of pyruvate oxidase (10, 11). The S. pneumoniae transparent and opaque variants differ in their ability to be transformed (20), and the transparent variants undergo premature lysis on agar surfaces relative to the opaque variants (19). The results that we have obtained with the S. pneumoniae type 2 strains containing either spxB or spxB⁺ are reminiscent of those that have been obtained with these opaque and transparent morphological variants. In fact, the colony morphology of spxB⁺-containing strains could be described as opaque and that of spxB-containing strains as transparent.

We have considered the possibility that Avery's group may have encountered opaque and transparent phase variation while working with R36 and that R36A was selected as a stable transparent variant. It is important to acknowledge that R36A and the modern-day transparent variants of others have been isolated with different growth conditions and their respective morphologies may not be the result of analogous biological processes. More information will be needed before any conclusions can be made about whether R36A is akin to the transparent variants that have already been described.

For clarity, we have confined our studies to R36A and its direct descendant R6, which was isolated from R36A sometime in the 1950s by Roland Hotchkiss (5, 14). Two widely used R36A descendants that were not considered here are the other common laboratory strains Rx1 and R6x. Rx1 is a spontaneous rough revertant of the R36A type 3 encapsulated strain S3-N (15) that integrates genetic markers into its genome with increased efficiency due to defects in mismatch repair (hex mutant) (17). R6x is a hex mutant that was derived by transforming R6 with genomic DNA from Rx1 (18). The available genetic and biochemical data suggest that Rx1 and R6x are not similar to R36A and R6 with regard to pyruvate oxidase. A comparison of the SpxB sequence of R6 with that of R6x (16) reveals that R6x SpxB does not contain the N95D mutation found in R36A and R6 SpxB but does contain the T282A mutation and five new amino acid substitutions. Work by others suggests that the alternative allele of spxB found in R6x may be associated with a decrease in pyruvate oxidase activity relative to that of R36A and R6. Hydrogen peroxide release assays with R6x and D39 reveal that the amount of hydrogen peroxide produced by R6x is comparable to that of D39 (13). In the same studies, Rx1 was found to produce only half as much hydrogen peroxide as D39 (13).

How and when Rx1 and R6x came to differ from R36A and R6 with regard to pyruvate oxidase activity, the relationship of pyruvate oxidase to the hex phenotype, and how the colony morphologies of R36A, R6, Rx1, and R6x relate to one another are all important questions that remain to be answered. That Rx1 and R6x may have sustained genetic changes that lower cellular pyruvate oxidase activity is perhaps not surprising. S. pneumoniae grows better when hydrogen peroxide levels are minimized, and genetic events altering the expression of pyruvate oxidase do appear to be favored in vitro. Mutation frequency in S. pneumoniae is influenced by hydrogen peroxide; the higher the levels of exogenous or endogenous hydrogen peroxide that cells are exposed to, the higher the mutation frequency (11). Thus, it might have been the high levels of hydrogen peroxide associated with the spxB allele that ultimately led to the changes in pyruvate oxidase activity in these other strains.